Next challenge: a machine that can can beat a human using an equal amount of energy.
I don't get the energy point. The machine has no health care costs and can play 24/7. Doesn't that count for something?
But your wish will come true. Go isn't a special snowflake. If you have an objective metric of success in a formal universe machines always win.
By your definition, on AI side we should add energy spent on creating AI and civilization that produced it.
Did... you really just say that, about a process that requires not only a civilization, but one sufficiently decadent that it can afford to waste resources making silicon that turns burned coal into pointless game victories?
The machine has no health care costs and can play 24/7.
I think by your previous metrics, you should be counting the heathcare costs of the ops folks who run the hardware, and I guess the healthcare costs of their healthcare workers, ad nauseum.
Ke Jie is currently 19, so he has used an estimate of 670kWH for both his training set and playing his games.
Getting a machine to win that consumes less than 97W/H would be hard.
Taking a new machine and loading training data on it and having it win would be far less than 670kWH
97W, not 97W/h.
Watts is already "per hour", mind. Or per second to be precise.
In my understanding, power (instantaneous) is measured in W, but consumption needs to be integrated over time, thus the per hour part.
The discussion is energy use at play time. For each given second, a certain number of joules are being used to compute a decision. That number as of today, in an unaided match, is independent of civilization's technological state.
That said, AlphaGo has seen a huge (10x?) gain in efficiency according to David Silver. Still far from a human but nonetheless very impressive drop in just a year.
But seriously, it's possible that AlphaGo is already much more energy efficient than a human player. The main reason it uses tons of energy, is the tree search part of the algorithm. Where it runs hundreds of thousands of simulated games to further analyze every move. This improves it's skill, but only by a little bit. IIRC, the version without tree search beat the full version 25% of the time. Which would still give it a higher elo than Sedol, which only beat it 20% of the time (and AlphaGo has improved since those games.)
Google is also using custom TPUs, which are claimed to be something like an order of magnitude or more energy efficient than GPUs. And computing technology is only getting more energy efficient with time. In principle, transistors moving around a few electrons are vastly more energy efficient than the very wasteful chemical reactions used in brain. We also know how to "sparsify" nets and remove tons unnecessary connections that could reduce computations a lot. But there's generally no point in doing that because it's not faster on normal hardware.
That would be amazing but it seems hard to believe. Any references?
I found this (which is also impressive):
AlphaGo team then tested the performance of the policy
networks. At each move, they chose the actions that were
predicted by the policy networks to give the highest
likelihood of a win. Using this strategy, each move took
only 3 ms to compute. They tested their best-performing
policy network against Pachi, the strongest open-source
Go program, and which relies on 100,000 simulations of
MCTS at each turn. AlphaGo's policy network won 85% of
the games against Pachi!
2. https://gogameguru.com/i/2016/03/deepmind-mastering-go.pdf
>In a similar matchup, AlphaGo running on multiple computers won all 500 games played against other Go programs, and 77% of games played against AlphaGo running on a single computer.
But the full version of AlphaGo that runs on thousands of computers is much stronger than that, so I was mistaken.
Still, the fact that the non-distributed version is so strong even without tree search is pretty amazing. It beat all existing Go playing programs a majority of the time. And with algorithmic advances and more training it may eventually catch up to best human players.
Non-distributed Alpha Go won 99% of the time versus just the value network and policy network with no rollouts. That AI was estimated as having a 2177 Elo rating, which is not very strong and much weaker than Sedol.
Even with a TPU, a human is more efficient. That neural net pair used 8 GPUs. At a generous 200 watt per GPU that's 1.6 kW, 10% of which is 160 watts. A human brain does all higher level reasoning and uses ~20 Watts. A human is not devoting 100% of its computational power on Go. It is likely just a fraction of that.
But if we look at Chess, Chess engines that run on mobile phones are possibly about or maybe slightly more efficient.
>In a similar matchup, AlphaGo running on multiple computers won all 500 games played against other Go programs, and 77% of games played against AlphaGo running on a single computer.
But you are right, the full version running on thousands of computers is much stronger than that.
Still, the fact that the non-distributed version is so strong even without tree search is pretty amazing. With algorithmic advances and more training it may eventually catch up to best human players. It's only the first generation of deep learning based Go bots.
And I believe the policy network only takes a few milliseconds to compute a move. So even if the TPU consumes hundreds of watts at full use, it doesn't need to run at full use for long.